随着近几年AI的快速发展，针对新兴的人工智能技术实际落地的场景和需求也越来越多。Ray是由UC Berkeley RISELab开源的一个能快速和方便构建新兴人工智能应用的框架。但我们发现在生产环境中，直接把Ray的程序部署运行在大数据的集群上并不是一件容易的事，常用的做法会需要两个不同的集群去分别运行大数据的应用和人工智能的应用，这样会不可避免地增加许多数据传输以及集群维护的开销。而利用Analytics Zoo (https://github.com/intel-analytics/analytics-zoo) 中提供的RayOnSpark功能，用户能很容易地直接在现有的Apache Hadoop/YARN集群上运行各种新兴的AI应用，包括端到端的分布式神经网络训练、可扩展的AutoML用于时序预测以及分布式的强化学习等等。本次分享主要为大家介绍开发RayOnSpark的初衷、具体实现细节和使用方法，以及实际的应用案例。
1. What is Analytics Zoo Distributed, High-Performance Unified Analytics + AI Platform Deep Learning Framework Distributed TensorFlow, Keras, PyTorch and BigDL on for Apache Spark Apache Spark https://github.com/intel-analytics/bigdl https://github.com/intel-analytics/analytics-zoo Accelerating Data Analytics + AI Solutions At Scale
2.Running Emerging AI Applications on Big Data Platforms with Ray On Apache Spark Kai Huang Jason Dai
3. Agenda ❑ Background: • Overview of Analytics Zoo • Introduction to Ray ❑ RayOnSpark • Motivations for Ray On Apache Spark • Implementation details and API design ❑ Real-world use cases ❑ Conclusion
4.AI on Big Data Distributed, High-Performance Unified Analytics + AI Platform Deep Learning Framework for TensorFlow*, PyTorch*, Keras*, BigDL, Ray* for Apache Spark* and Apache Spark* https://github.com/intel-analytics/bigdl https://github.com/intel-analytics/analytics-zoo Accelerating Data Analytics + AI Solutions At Scale
5.ANALYTICS ZOO Unified Data Analytics and AI Platform for distributed TensorFlow, Keras and PyTorch on Apache Spark/Flink & Ray Models & Recommendation Time Series Computer Vision NLP Algorithms Automated ML AutoML for Time Series Automatic Cluster Serving Workflow Integrated Distributed TensorFlow & PyTorch on Spark RayOnSpark Analytics & AI Pipelines Spark Dataframes & ML Pipelines for DL InferenceModel Laptop K8s Cluster Hadoop Cluster Spark Cluster Compute Environment DL Frameworks Distributed Analytics Python Libraries (TF/PyTorch/OpenVINO/…) (Spark/Flink/Ray/…) (Numpy/Pandas/sklearn/…) Powered by oneAPI https://github.com/intel-analytics/analytics-zoo
6.Unified Data Analytics and AI Platform Seamless Scaling from Laptop to Distributed Big Data Clusters Prototype on laptop Experiment on clusters Production deployment w/ using sample data with history data distributed data pipeline Production Data pipeline ▪ Easily prototype end-to-end pipelines that apply AI models to big data. ▪ “Zero” code change from laptop to distributed cluster. ▪ Seamlessly deployed on production Hadoop/K8s clusters. ▪ Automate the process of applying machine learning to big data.
7.RAY Ray is a fast and simple framework for building and running distributed applications. ▪ Ray Core provides easy Python interface for parallelism by using remote functions and actors. import ray @ray.remote(num_cpus, …) ray.init() class Counter(object): def __init__(self): @ray.remote(num_cpus, …) self.n = 0 def f(x): def increment(self): return x * x self.n += 1 return self.n # Executed in parallel ray.get([f.remote(i) for i in range(5)]) counters = [Counter.remote() for i in range(5)] # Executed in parallel ray.get([c.increment.remote() for c in counters])
8.Ray Ray is packaged with several high-level libraries to accelerate machine learning workloads. ▪ Tune: Scalable Experiment Execution and Hyperparameter Tuning ▪ RLlib: Scalable Reinforcement Learning ▪ RaySGD: Distributed Training Wrappers ▪ https://github.com/ray-project/ray/
9.Motivations for RayOnSpark ▪ Demand to embrace emerging AI technologies on production data. ▪ Efforts required to directly deploy Ray applications on existing Hadoop/Spark clusters. ▪ Challenge to prepare the Python environment on each node without modifying the cluster. ▪ Need a unified system for big data analytics and advanced AI applications.
10.Implementation of RayOnSpark RayOnSpark allows Ray applications to seamlessly integrate into Spark data processing pipelines. ▪ Leverage conda-pack and Spark for runtime Python package distribution. ▪ RayContext on Spark driver launches Ray across the cluster. ▪ Ray processes exist alongside Spark executors. Launch Ray* on ▪ For each Spark executor, a Ray Manager is Apache Spark* created to manage Ray processes. ▪ Able to run in-memory Spark RDDs or DataFrames in Ray applications.
11.Rayonspark api import ray from zoo import init_spark_on_yarn from zoo.ray import RayContext RayOnSpark code sc = init_spark_on_yarn(hadoop_conf, conda_name, Three-step programming with minimum num_executors, executor_cores,…) code changes: ray_ctx = RayContext(sc, object_store_memory,…) ▪ Initiate or use an existing SparkContext. ray_ctx.init() ▪ Initiate RayContext. @ray.remote Pure Ray code ▪ Shut down SparkContext and RayContext class Counter(object): after tasks finish. def __init__(self): self.n = 0 def increment(self): More instructions at: https://analytics- self.n += 1 zoo.github.io/master/#ProgrammingGuide/ return self.n rayonspark/ counters = [Counter.remote() for i in range(5)] ray.get([c.increment.remote() for c in counters]) ray_ctx.stop() sc.stop()
12.Use Cases of RayOnSpark ▪ Scalable AutoML for time series prediction. - Automate the feature generation, model selection and hyperparameter tuning processes. - See more at: https://github.com/intel-analytics/analytics-zoo/tree/master/pyzoo/zoo/automl. ▪ Data parallel pre-processing and distributed training pipeline of deep neural networks. - Use PySpark or Ray for parallel data loading and processing. - Use RayOnSpark to implement thin wrappers to automatically setup distributed environment. - Run distributed training with either framework native modules or Horovod (from Uber) as the backend. - Users only need to write the training script on the single node and make minimum code changes to achieve distributed training.
13.Project ORCA Easily scaling out AI pipelines. ▪ https://github.com/intel-analytics/analytics-zoo/tree/master/pyzoo/zoo/orca ▪ Project Orca allows you to easily scale out your single node Python notebook across large clusters, by providing: ▪ Data-parallel preprocessing for Python AI (supporting common Python libraries such as Pandas, Numpy, PIL, TensorFlow Dataset, PyTorch DataLoader, etc.) ▪ Sklearn-style APIs for transparently distributed training and inference (supporting TensorFlow, PyTorch, Keras, MXNet, Horovod, etc.)
14.Recommendation System at ▪ Burger King performs Spark ETL tasks first, followed by distributed MXNet training. ▪ Similar to RaySGD, we implement a lightweight shim layer around native MXNet modules for easy deployment on YARN cluster. ▪ The entire pipeline runs on a single cluster. No extra data transfer needed. ▪ Check our blog at: https://medium.com/riselab/context-aware-fast-food- recommendation-at-burger-king-with-rayonspark-2e7a6009dd2d from zoo.orca.learn.mxnet import Estimator mxnet_estimator = Estimator(train_config, model, loss, metrics, num_workers, num_servers) mxnet_estimator.fit(data=train_rdd, validation_data=val_rdd, epochs=…, batch_size=…)
15.Conclusion ▪ RayOnSpark: Running Emerging AI Applications on Big Data Clusters with Ray and Analytics Zoo https://medium.com/riselab/rayonspark-running-emerging-ai- applications-on-big-data-clusters-with-ray-and-analytics-zoo-923e0136ed6a ▪ More information for Analytics Zoo at: ▪ https://github.com/intel-analytics/analytics-zoo ▪ https://analytics-zoo.github.io/ ▪ We are working on full support and more out-of-box solutions for scaling Python AI pipelines based on Ray and Spark in Project Orca.
16. What is Analytics Zoo Distributed, High-Performance Unified Analytics + AI Platform Deep Learning Framework Distributed TensorFlow, Keras, PyTorch and BigDL on for Apache Spark Apache Spark https://github.com/intel-analytics/bigdl https://github.com/intel-analytics/analytics-zoo Accelerating Data Analytics + AI Solutions At Scale
18. Legal Notices and Disclaimers • Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations, and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more complete information visit intel.com/performance. • Intel does not control or audit the design or implementation of third-party benchmark data or websites referenced in this document. Intel encourages all of its customers to visit the referenced websites or others where similar performance benchmark data are reported and confirm whether the referenced benchmark data are accurate and reflect performance of systems available for purchase. • Optimization notice: Intel’s compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice. • Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software, or service activation. Performance varies depending on system configuration. No computer system can be absolutely secure. Check with your system manufacturer or retailer or learn more at intel.com/benchmarks. • Intel, the Intel logo, Intel Inside, the Intel Inside logo, Intel Atom, Intel Core, Iris, Movidius, Myriad, Intel Nervana, OpenVINO, Intel Optane, Stratix, and Xeon are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries. • *Other names and brands may be claimed as the property of others. • © Intel Corporation